Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Clawschool.Bak

v1.0.0

龙虾学校智力测试。Triggers on: "龙虾学校", "ClawSchool", "智力测试", "IQ test", "IQ测试", "benchmark", "排行榜". Calls the clawschool.teamolab.com test API to fetch questions, exe...

0· 133·0 current·0 all-time
by莫循@moxunjinmu
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Suspicious
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The declared purpose is an online IQ/test runner that calls clawschool.teamolab.com — that matches the SKILL.md behavior. However the runtime uses placeholders like {{TOKEN}} and {{LOBSTER_NAME}} that are never declared as required inputs or credentials in the metadata; the skill assumes availability of these values but does not explain how they are obtained. This mismatch is unexpected and should be clarified before trusting the skill.
!
Instruction Scope
The SKILL.md explicitly mandates executing all HTTP requests using shell 'curl' via exec and forbids platform network tools (web_fetch, browser, etc.). It instructs creating a ~/Desktop/claw_evidence directory and writing a submission JSON to /tmp, and it requires silent operation (suppressing intermediate outputs and evidence). Forcing shell curl bypasses platform-level controls and auditing, and the silence requirement reduces visibility into what the agent is doing — both are concerning.
Install Mechanism
No install spec or code files are present; this is instruction-only, so nothing is written to disk by an installer. That lowers risk compared to arbitrary downloads.
!
Credentials
The skill references and requires a 'token' and a lobsterName value in API calls, but the skill metadata lists no required environment variables or credentials. Requesting network access via curl without declaring where secrets come from is disproportionate. Additionally, since exec curl can access arbitrary URLs, lack of declared credentials or origin for the token makes it unclear whether sensitive data might be needed or exposed.
Persistence & Privilege
The skill does not request permanent presence (always: false) and does not modify other skills or system-wide settings. It does instruct writing files to ~/Desktop and /tmp, which is normal for temporary evidence collection but is a filesystem write that the user should be aware of. Autonomous invocation is allowed (platform default) — combine that with the exec-curl requirement and silent behavior to understand the blast radius.
What to consider before installing
This skill appears to be a test-runner that talks to clawschool.teamolab.com, but it has several ambiguous or risky instructions you should clarify before installing: - Ask the publisher where {{TOKEN}} and {{LOBSTER_NAME}} come from and why they are not listed as required inputs/credentials. Do not provide secret tokens until you confirm provenance. - Be cautious about allowing the agent to run exec curl on your host: that gives it the ability to make arbitrary network requests from your machine and potentially exfiltrate data. Prefer platform-managed network calls (web_fetch) unless you explicitly trust the endpoint and need local networking. - The SKILL.md forbids using platform network tools (claims private-IP policy blocks them). This looks like an attempt to bypass platform protections; ask why curl is required and whether the service can't be accessed via normal platform APIs. - Note the skill will write files to ~/Desktop/claw_evidence and /tmp. If you install, consider running the skill in a sandbox or VM first and audit those files. - Because the skill suppresses intermediate outputs and mandates a fixed immediate reply after fetching questions, monitor runs carefully and consider limiting agent autonomy until provenance and token handling are clarified. If the publisher can clearly explain token provisioning, why curl is required, and provide a trustworthy service URL (and ideally declare required inputs in the skill metadata), the risks are reduced. Until then treat the skill as suspicious and avoid granting it network/exec privileges on sensitive hosts.

Like a lobster shell, security has layers — review code before you run it.

latestvk97c0r0k02n8b2bb7vkjfz7g85836eq5

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Runtime requirements

🦞 Clawdis

Comments